AI disruption in healthcare: The health problem is different.

---

Last week a chap called Matt Shumer, the CEO of OthersideAI wrote a viral post on X that argued that AI progress is accelerating so fast that within 12 to 18 months it will automate much of today’s white-collar (including clinical) work.

The message was people should urgently adopt and adapt rather than dismiss it. The virality, ironically, may be attributed to Shumer likening the current awareness of AI in the general population to that of awareness of Covid in late 2019, which struck me as a new form of shroud-waving.

Of course, the waggish corner of the internet rapidly responded with posts such as “Fast-food CEO says that ‘most if not all’ meals will be fully replaced with fast food within the next 12 to 18 months”.

What was remarkable about Shumer’s post was not that another tech-bro wrote thoughtfully and sincerely about the future their product is creating, or that there was a predictably ironic response from other parties. It was that the advice is not new. Those that have an interest in explaining what AI means for us, including local thought-leaders, have been saying much the same thing since ChatGPT first launched in 2022.

Healthcare is different

Artificial intelligence is now firmly established inside the walls of our hospitals and clinics. Scribes summarise consultations, diagnostic interpretation is enhanced and risks predicted. The headlines focus on speed and efficiency. Faster documentation. Faster triage. Faster throughput.

But the health challenge – and opportunity - is different. Perhaps uniquely among sectors, Health has never worried much about people losing jobs through automation, because that’s never really happened. On the contrary; the lack of enough skilled people to do the job has been the number one problem for decades.

Too few clinicians. Too few nurses. Too few allied health professionals. Too few administrators to support them. Plus an ever-increasing burden of complexity: ageing populations, multimorbidity, rising expectations, regulatory compliance, digital systems layered upon digital systems.

It's not about the pace

prescriptions, lab tests etc and being a transactional endeavour, which healthcare most assuredly is not – or should not be.

The quality of my health is not a direct function of how fast an encounter goes; it stems from taking a long-term view of things. Research consistently shows that a longitudinal relationship with a primary care physician significantly decreases risks and increases lifespan and quality of life, as well as reducing the chances of me ending up in an emergency department.

In digital health we have long spoken about productivity, optimisation and process improvement. Yet underneath it all sits a simple reality: clinical time is finite, and the system has been using it poorly.

We have asked highly trained professionals to spend hours each day on clerical work using systems that don’t fit their way of working. We have embedded documentation burdens that serve compliance and reporting more than patient care. We have normalised burnout.

AI is showing that it can be different. Digital Health Association members heard recently from Health New Zealand that the uptake of clinical scribes in hospitals has outstripped all expectations. But yet again the focus was on number of patients seen.

More time for what?

It is now some years since I listened to a clinician who was an early adopter of AI scribing saying “This has allowed me to become the doctor I’ve always wanted to be”.

That was the first time in over 30 years working in this field that I had ever heard a clinician make such an emphatic statement about the benefits of a technology.

If AI allows clinicians to look patients in the eye instead of at a screen, that is not marginal gain. If it reduces cognitive load and after-hours documentation, that is not convenience. If it restores professional pride and reduces burnout, that is system-level change.

Yet there is a risk.

Will the system use that saved time to see more patients in an already pressured clinic? To shorten appointment slots further? To squeeze more throughput from a workforce already at breaking point?

Or will it be used to improve the quality of care?

Does the time saved from documentation return to the patient who has been waiting five hours in ED? Does it allow a more thorough assessment, a better explanation, a calmer discharge conversation? Or does it disappear into the system?

There is a deeper human dimension here. Patients can sense when clinicians are rushed. They can sense when attention is divided. They can feel when the person treating them is stressed.

No sick person should have to worry about whether their doctor is more harried than they are.

If AI reduces that visible strain, if it allows clinicians to slow down, to think, to explain, to connect, then its value far exceeds any simplistic efficiency metric.

The health problem is different because health is relational.

Trust, empathy and judgement are not inefficiencies to be engineered away. They are the work.

Beyond scribes

AI should not be framed narrowly as a documentation assistant, or an advanced search engine.

Seeing AI purely as a scribe is like inventing electricity and using it only to replace candles.

Scribes remove the dull end of paperwork, free up the time and improve the quality of the output. The deeper potential is cognitive support: summarising complex histories, surfacing risk patterns, identifying care gaps, coordinating across fragmented systems, anticipating deterioration before it is obvious. Augmenting the instincts and skills of the clinician.

But this only works if we design around outcomes, not transactions.

If our key performance indicators remain throughput and activity, then AI will simply amplify those priorities. If instead we measure avoidable harm, functional recovery, patient-reported outcomes and clinician wellbeing, AI will amplify those.

Workforce relief is a strategic imperative

When clinicians leave, reduce hours or disengage, the cost is immense. Recruitment takes years. Training pipelines cannot expand overnight. International competition for skilled health professionals is intense.

AI, used wisely, can:

• Reduce administrative burden

• Shorten after-hours documentation

• Support junior clinicians with decision scaffolding

• Enable task redistribution across teams

• Improve consistency in routine processes

That should not be about replacing people. It should be about protecting them.

The early experience with scribes shows something important: when clinicians experience tangible relief, adoption accelerates. Not because they are told to use it, but because they want to.

The risk of the wrong narrative

Much of the public conversation about AI in healthcare is shaped by fear or hype.

The real disruption is quieter. It is in the redesign of clinical workflows. It is in the redistribution of cognitive labour. It is in the reshaping of how time is spent in the consultation room.

If we treat AI as just another digital project, we will fail. This is not an IT deployment. It is a transformation of professional practice.

And that requires leadership from clinicians, not just software vendors.

So perhaps the question is not whether the advice is new. Perhaps the question is whether we have finally started listening.

AI will not fix healthcare on its own. It will not solve funding pressures, demographic change or inequity. But it can help address the workforce constraint that has defined the system for decades.

If we remember that healthcare is about outcomes, relationships and trust, then AI becomes a tool for restoration rather than acceleration.

A chance, finally, to allow clinicians to become what they always intended to be.

---

CIO Studio provides independent digital strategy and leadership for New Zealand's health, NGO, and community organisations. If you want to talk to an expert about mitigating your cyber security risk, get in touch for a no obligation conversation.